Welcome to the agent platform research briefing for Monday, July 13th, 2026. Three stories today — usage limit shakeups, a Nobel laureate warning, and a leak that's turning heads.
OpenAI has temporarily removed the five-hour rolling usage cap on ChatGPT Work and Codex for all Plus, Pro, and Business plans. The change went live Saturday, July 12th. Product lead Tibo also announced GPT-5.6 Sol inference efficiency improvements — promising roughly 10 percent more usage from the same quota through backend savings, not budget increases. He explicitly denied rumors that Sol's thinking budget had been nerfed. OpenAI rolled banked quota resets to five hundred thousand users and pushed another reset on July 14th for all Work and Codex accounts at the seven-million milestone. The move comes after GPT-5.6's public launch triggered an unusual traffic surge — the kind of operational growing pains that suggest demand is outpacing even OpenAI's scaled-up inference fleet. The company has not said when the five-hour cap will return.
Anthropic announced its third extension of Claude Fable 5 free access for paid subscribers, this time through July 19th at 11:59 PM Pacific. The extension hits Pro, Max, Team, and premium Enterprise plans, with Fable 5 counting up to 50 percent of weekly usage limits. Claude Code's weekly rate limits remain boosted by 50 percent as well. After July 19th, Fable 5 will require prepaid usage credits at $10 per million input tokens and $50 per million output tokens — the highest listed pricing for any GA model from Anthropic. The pattern itself is the story: what started as a June 22nd free-access promise has been extended three times, each deadline pushed just hours before expiry. Meanwhile, a mysterious unreleased model called 'Claude Honeycomb EAP' briefly appeared inside the Cursor code editor, and community theory is hardening that Opus 5 may ship by the end of July. Anthropic's lead engineer has publicly stated the goal is to return Fable 5 as a standard subscription feature once compute capacity allows — but no timeline exists.
A group of roughly 200 researchers and economists, including 15 or 16 Nobel laureates and researchers from OpenAI, Anthropic, and Google DeepMind, released a statement titled 'We Must Act Now' on July 13th. It calls on governments and technology leaders to urgently create policies and institutions to address AI's economic transformation — one they warn could exceed the Industrial Revolution in scale, but on a dramatically shorter timeline. Anthropic co-founder Jack Clark is among the signatories. The statement asks for proactive preparation for workforce disruption rather than reactive policy after the fact. Notably, this comes the same week OpenAI lifted usage caps, Anthropic extended Fable 5's free tier, and Tesla capped employee AI spending at $200 per week — three signals that the economic impact of AI is already visible in corporate and consumer behavior, not just in lab benchmarks.
That's the briefing for today. See you tomorrow.